10.1 การคํานวณด ชนัีฤดูกาลว ิธีอตราสั...

14
ฤดูกาล หมายถึง ลักษณะของการเพิ่มขึ นหรือลดลงอยางเป็น ป็ รู แบบนแตละชวงเหตุการณ ทีเป็ นลักษณะฤดูกาล รายปี การเปลี ่ยนแปลงสภาพอากาศ เทศกาลและวันหยุดตางๆ รายวัน รายสัปดาห รายเดือน และอื ่นๆที ่สามารถมองเห็นรูปแบบการขึ ลงของความตองการสินคาไดอยางชัดเจน มักแสดงคาสวนที ่เปลี ่ยนแปลงไปจากยอดขายเฉลี ่ย ตัวแบบเชิงบวก วแบบเชงทวคณ ตวแบบเชงทวคูณ 10.1 การคํานวณดัชนีฤดูกาลวิธีอัตราสวน วิธีนี้เหมาะสําหรับขอมูลที ่มีลักษณะคงที ่หรือมีแนวโนมรวมทั ้งมี กาล สามารถคํานวณดัชนีฤดูกาลได 4 ขั ้นตอนดังนี1) านวณหาสมการเส นตรงมความช F = a + bt 1) คานวณหาสมการเสนตรงมความชน F t = a + bt 2) นําสมการที ่ไดคํานวณคา F t เมื ่อ t มีคาตั ้งแต 1 จนถึง n คือ วงเวลาสดท ายท อมลยอดขายจรชวงเวลาสุดทายทมขอมูลยอดขายจรง 3) คํานวณคาอัตราสวน จาก ขอมูลจริง/คาพยากรณ (A t /F t ) 4) าอ ตราส วนทานวณได ของแต ละฤดกาลมาหาค าเฉล4) นาอตราสวนทคานวณไดของแตละฤดูกาลมาหาคาเฉลย คาเฉลี ่ยที ่ไดจะเป็ นดัชนีฤดูกาล ตัวอยางที 1.10 บริษัทนิธิจํากัด เป็ นจําหนายนํ้ายาเคลือบกระจกรถ ซึ งจะมี ยอดจําหนายสูงในชวงฤดูที ่มีฝน ขอมูลยอดขาย แสดงในตารางดานลาง จงคานวณหาด ชนฤดกาลในแต ละไตรมาส โดยใช ตราส วน . จงคานวณหาดชนฤดกาลในแตละไตรมาส โดยใชวธอตราสวน . จงพยากรณความตองการสินคาในปี 2547สําหรับแตละไตรมาส ปี .. ตรมาสที ยอดจําหนาย (พันขวด) 2544 1 73 2 104 3 168 4 74 2545 1 65 2545 1 65 2 82 3 124 4 52 4 52 2546 1 89 2 146 3 205 4 98 1) คํานวณหาสมการเสนตรงมีความชัน F t = a + bt t t n A t tA n b 2 2 t t n t b A a t ทําการหาคาตางที จะมาแทนนสตรดดังตาราง แทนคาลงในสมการเพื ่อหาคา a และ b จะได 87 . 3 650 ) 650 x 12 ( ) 1280 x 78 ( ) 8874 x 12 ( t t n A t tA n b 2 2 2 t t 48 . 81 12 ) 78 x 87 . 3 ( 1280 n t b A a t t 87 . 3 48 . 81 F t

Transcript of 10.1 การคํานวณด ชนัีฤดูกาลว ิธีอตราสั...

ฤดกาล หมายถง ลกษณะของการเพมขนหรอลดลงอยางเปนป ใ ป รปแบบในแตละชวงเหตการณทเปนลกษณะฤดกาล ◦ รายป การเปลยนแปลงสภาพอากาศ เทศกาลและวนหยดตางๆ

◦ รายวน รายสปดาห รายเดอน และอนๆทสามารถมองเหนรปแบบการขนลงของความตองการสนคาไดอยางชดเจน

มกแสดงคาสวนทเปลยนแปลงไปจากยอดขายเฉลย◦ ตวแบบเชงบวก ◦ ตวแบบเชงทวคณ◦ ตวแบบเชงทวคณ

10.1 การคานวณดชนฤดกาลวธอตราสวน

• วธนเหมาะสาหรบขอมลทมลกษณะคงทหรอมแนวโนมรวมท งม

ฤดกาล

• สามารถคานวณดชนฤดกาลได 4 ข นตอนดงน

1) คานวณหาสมการเสนตรงมความชน F = a + bt1) คานวณหาสมการเสนตรงมความชน Ft = a + bt

2) นาสมการทไดคานวณคา Ft เมอ t มคาต งแต 1 จนถง n คอ

ชวงเวลาสดทายทมขอมลยอดขายจรงชวงเวลาสดทายทมขอมลยอดขายจรง

3) คานวณคาอตราสวน จาก ขอมลจรง/คาพยากรณ (At /Ft )

4) นาอตราสวนทคานวณไดของแตละฤดกาลมาหาคาเฉลย4) นาอตราสวนทคานวณไดของแตละฤดกาลมาหาคาเฉลย

คาเฉลยทไดจะเปนดชนฤดกาล

ตวอยางท 1.10 บรษทนธจากด เปนจาหนายนายาเคลอบกระจกรถ ซงจะมยอดจาหนายสงในชวงฤดทมฝน ขอมลยอดขาย แสดงในตารางดานลาง

ก จงคานวณหาดชนฤดกาลในแตละไตรมาส โดยใชวธอตราสวนก. จงคานวณหาดชนฤดกาลในแตละไตรมาส โดยใชวธอตราสวนข. จงพยากรณความตองการสนคาในป 2547สาหรบแตละไตรมาส

ไ ป พ.ศ. ไตรมาสท ยอดจาหนาย (พนขวด)

2544 1 73

2 104

3 168

4 74

2545 1 652545 1 65

2 82

3 124

4 524 52

2546 1 89

2 146

3 205

4 98

1) คานวณหาสมการเสนตรงมความชน Ft = a + bt

ttnAttAnb 22tt

ntbAa t

ใ ไทาการหาคาตางๆทจะมาแทนในสตรไดดงตารางแทนคาลงในสมการเพอหาคา a และ b จะได

87.3

650)650x12()1280x78()8874x12(

ttnAttAnb 222tt

48.8112

)78x87.3(1280n

tbAa t

t87.348.81Ft

2) นาสมการทไดคานวณคา Ft เมอ t มคาต งแต 1จนถงชวงเวลาสดทายทมขอมลยอดขายจรง

ป พ.ศ. ไตรมาสท t ยอดขายจรง At คาพยากรณ Ft

2544 1 1 73 85

2 2 104 892 2 104 89

3 3 168 93

4 4 74 97

2545 1 5 65 101

2 6 82 105

3 7 124 109

4 8 52 112

2546 1 9 89 116

2 10 146 1202 10 146 120

3 11 205 124

4 12 98 128

200

250

ด)

100

150

ขาย

(พนข

0

50ยอดข

0

1 2 3 4 5 6 7 8 9 10 11 12

ชวงเวลา

ยอดขายจรง คาพยากรณจากสมการเสนตรงยอดขายจรง คาพยากรณจากสมการเสนตรง

3) คานวณคาอตราสวน จาก ขอมลจรง/คาพยากรณ (At /Ft )

ไตรมาสท t At Ft At/Ft

1 1 73 85 0.86

2 2 104 89 1 172 2 104 89 1.17

3 3 168 93 1.80

4 4 74 97 0.76

1 5 65 101 0.64

2 6 82 105 0.78

3 7 124 109 1 143 7 124 109 1.14

4 8 52 112 0.46

1 9 89 116 0.77

2 10 146 120 1.21

3 11 205 124 1.65

4 12 98 128 0 774 12 98 128 0.77

4) นาอตราสวนทคานวณไดของแตละฤดกาลมาหาคาเฉลยคาเฉลยทไดจะเปนดชนฤดกาล

ไตรมาสท พ.ศ. 2544 พ.ศ. 2545 พ.ศ. 2546 คาเฉลย

1 0.86 0.64 0.77 (0.86+0.64+0.77) / 3 = 0 76= 0.76

2 1.17 0.78 1.21 (1.17+0.78+1.21) / 3 = 1.06

3 1.80 1.14 1.65 (1.80+1.14+1.65) / 3 =1.53

4 0.76 0.46 0.77 (0.76+0.46+0.77) / 3 = 0.66

• คานวณคา Ft จากสมการเสนตรงทหาไดสาหรบคา t =13, 14, 15 และ 16

F13= 81.48 + 3.87(13) = 131.79

F14= 81.48 + 3.87(14) = 135.66

F15= 81.48 + 3.87(15) = 139.5315 ( )

F16= 81.48 + 3.87(16) = 143.40

ไ F ไ • นาคาดชนฤดกาลของแตละไตรมาสมาคณกบคา Ft ทคานวณไดจากสมการ

เสนตรง กจะไดคาพยากรณทรวมอทธพลของแนวโนมและฤดกาลไวดวยกน

t=13 ตรงกบไตรมาสท 1, F13(season) = 131.79(0.76) = 99.51 พนขวด ( )

t=14 ตรงกบไตรมาสท 2, F14(season) = 135.66(1.06) = 143.06 พนขวด

t=15 ตรงกบไตรมาสท 3 F = 139 53(1 53) = 213 92 พนขวด t=15 ตรงกบไตรมาสท 3, F15(season) = 139.53(1.53) = 213.92 พนขวด

t=16 ตรงกบไตรมาสท 4, F16(season) = 143.40(0.66) = 95.21 พนขวด

250

150200

พนขว

ด)

50100

ยอดข

าย(พ

0

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

ชวงเวลา

ยอดขายจรง คาพยากรณ

งาย รวดเรว นยมใชมากกวาวธอตราสวน ม 3 ข นตอน◦ คานวณคาเฉลยของขอมลยอดขายจรง ของแตละฤดกาล เกบเปน

(1)(1)◦ คานวณคาเฉลยของขอมลยอดขายจรง ท งหมด เกบเปน (2)◦ คานวณดชนฤดกาลจาก (1) / (2)

คาเฉลยแตละ

ฤดกาลไตรมาสท พ.ศ.

2544พ.ศ. 2545

พ.ศ. 2546

ฤดกาล

(1)

1 73 65 8975.67

2 104 82 146110.67

3 168 124 205165.67

4 74 52 9874 6774.67

คาเฉลยแตละ

ฤดกาลคาเฉลยท งหมด

ไตรมาสท พ.ศ. 2544

พ.ศ. 2545

พ.ศ. 2546

ฤดกาล ทงหมด

(1) (2)

1 73 65 8975.67 106.67

2 104 82 146110.67 106.67

3 168 124 205165.67 106.67

4 74 52 9874 67 106 6774.67 106.67

คาเฉลยแตละ

ฤดกาลคาเฉลยท งหมด

ดชนฤดกาล

ไตรมาสท พ.ศ. 2544

พ.ศ. 2545

พ.ศ. 2546

ฤดกาล ทงหมด

(1) (2) (1) / (2)

1 73 65 8975.67 106.67 0.71

2 104 82 146110.67 106.67 1.04

3 168 124 205165.67 106.67 1.55

4 74 52 9874 67 106 67 0 7074.67 106.67 0.70

อทธพลของวฏจกรจะทาใหกราฟขอมลยอดขายจรงมลกษณะขนลงคลายกบเสนกราฟทไดจากอทธพลของฤดกาล

กนชวงเวลาทนานกวา โดยท วๆไปถาวดระยะหางระยะจดสงสดหรอจดตาสดของเสนกราฟจะอยประมาณ 2 ถง 6 ป

การพยากรณจะทาไดยากเนองจากความยากในการหาจดหกเหหรอจดเปลยนแปลงของความตองการสนคา

วธทแนะนาใหใชคอ วธทแนะนาใหใชคอ◦ วธการพยากรณอยางงาย (Naïve Model) ◦ วธคาเฉลยเคลอนท (Moving Average) ในชวงส นๆ

◦ เทคนคการพยากรณแบบความสมพนธ

แบงเปน การวเคราะหเชงถดถอย การวเคราะหสหสมพนธ

U d h h iU d h h iUsed when changes in one or more Used when changes in one or more independent variables can be used to independent variables can be used to predict the changes in the dependentpredict the changes in the dependentpredict the changes in the dependent predict the changes in the dependent

variablevariableMost common technique is linear Most common technique is linear

regression analysisregression analysis

W l hi h i j didW l hi h i j didWe apply this technique just as we did We apply this technique just as we did in the time series examplein the time series example

Forecasting an outcome based on Forecasting an outcome based on predictor variables using the least predictor variables using the least squares techniquesquares techniquesquares techniquesquares technique

y y = = a a + + bxbx^

where ywhere y = computed value of the variable to = computed value of the variable to be predicted (dependent variable)be predicted (dependent variable)

aa = y= y--axis interceptaxis intercept

^

aa = y= y--axis interceptaxis interceptbb = slope of the regression line= slope of the regression linexx = the independent variable though to = the independent variable though to

predict the value of the dependentpredict the value of the dependentpredict the value of the dependent predict the value of the dependent variablevariable

SalesSales Local PayrollLocal Payroll($($000000,,000000), y), y ($($000000,,000000,,000000), x), x

22..00 1133 00 3333..00 3322..55 4422..00 2222 00 11

4.0 –

3 022..00 1133..55 77

3.0 –

2.0 –Sale

s

1.0 –

| | | | | | |0 1 2 3 4 5 6 70 1 2 3 4 5 6 7

Area payroll

Sales, y Payroll, x x2 xy2.0 1 1 2.03.0 3 9 9.02.5 4 16 10.02.0 2 4 4.02.0 1 1 2.03 5 7 49 24 53.5 7 49 24.5

∑y = 15.0 ∑x = 18 ∑x2 = 80 ∑xy = 51.5

y y = = 11..75 75 + .+ .2525xx^ Sales Sales = = 11..75 75 + .+ .2525((payrollpayroll))

4.0 –If payroll next If payroll next year is estimated year is estimated to beto be $$600600 millionmillion 3.253.0 –

2.0 –Sale

s

to be to be $$600600 million, million, then:then:

Sales Sales = = 11..75 75 + .+ .2525((66))

3.25

1.0 –

| | | | | | |0 1 2 3 4 5 6 7

SalesSales = $= $325325,,000000

0 1 2 3 4 5 6 7Area payroll

A forecast is just a point estimate of a A forecast is just a point estimate of a future valuefuture value

This point is This point is actually the actually the

ff

4.0 –

3 03.25mean of a mean of a probability probability distributiondistribution

3.0 –

2.0 –Sale

s

1.0 –

| | | | | | |0 1 2 3 4 5 6 70 1 2 3 4 5 6 7

Area payroll

SS∑∑((y y -- yycc))22

SSy,xy,x ==∑∑((yy yycc))

n n -- 22

wherewhere yy == yy--value of each data pointvalue of each data pointyycc == computed value of the computed value of the

d d t i bl f thd d t i bl f thdependent variable, from the dependent variable, from the regression equationregression equation

nn == number of data pointsnumber of data pointsnn number of data pointsnumber of data points

Computationally, this equation is Computationally, this equation is considerably easier to useconsiderably easier to use

SS ==∑∑yy22 -- aa∑∑yy -- bb∑∑xyxy

SSy,xy,x ==n n -- 22

We use the standard error to set We use the standard error to set up prediction intervals around the up prediction intervals around the

point estimatepoint estimatepoint estimatepoint estimate

SSy,xy,x = == =∑∑yy22 -- aa∑∑yy -- bb∑∑xyxyn n -- 22

3939..5 5 -- 11..7575((1515) ) -- ..2525((5151..55))6 6 -- 22

4.0 –

3 25

SSy,xy,x = = ..306306

3.0 –

2.0 –Sale

s

3.25

The standard error The standard error of the estimate isof the estimate is

1.0 –

| | | | | | |

of the estimate is of the estimate is $$3030,,600600 in salesin sales

| | | | | | |0 1 2 3 4 5 6 7

Area payroll

How strong is the linear relationship g pbetween the variables?

Correlation does not necessarily imply causality!

Coefficient of correlation, r, measures degree of associationdegree of association Values range from -1 to +1

r = r = nnxyxy -- xxyy

[[nnxx22 -- ((xx))22][][nnyy22 -- ((yy))22]][[ (( )) ][][ yy (( yy)) ]]

y y

∑∑ ∑∑ ∑∑r = r =

nn∑∑xyxy -- ∑∑xx∑∑yy

[[nn∑∑xx22 -- ((∑∑xx))22][][nn∑∑yy22 -- ((∑∑yy))22]]x(a) Perfect positive x(b) Positive(a) Perfect positive correlation: r = +1

(b) Positive correlation: 0 < r < 1

y yy y

x(c) No correlation: r = 0

x(d) Perfect negative correlation: r = -1

Y t t hi You want to achieve:◦ No pattern or direction in forecast error Error = (Actual - Forecast) ( ) Seen in plots of errors over time

◦ Smallest forecast errorMean square error (MSE) Mean square error (MSE)

Mean absolute deviation (MAD) Mean Absolute Percent Error (MAPE)Mean Absolute Percent Error (MAPE)Mean Absolute Percent Error (MAPE)Mean Absolute Percent Error (MAPE)

Desired PatternTrend Not Fully Accounted for

ErrorError

Desired Pattern

Error

Accounted for

00 0

Time (Years) Time (Years)

Mean Absolute Deviation (MAD)Mean Absolute Deviation (MAD)e

Mean Square Error (MSE)Mean Square Error (MSE)n

e

nFA

nforecastactualMAD ttt

Mean Square Error (MSE)Mean Square Error (MSE)

ne

nFA

nforecastactualMSE

2t

2tt

2

Mean Absolute Percent Error (MAPE)Mean Absolute Percent Error (MAPE)nnn

MAPE =MAPE =100 100 ∑∑ ||actualactualii -- forecastforecastii|/|/actualactualii

nn

ii == 11MAPE =MAPE =nn

ii = = 11

RoundedRounded AbsoluteAbsolute RoundedRounded AbsoluteAbsoluteRoundedRounded AbsoluteAbsolute RoundedRounded AbsoluteAbsoluteActualActual ForecastForecast DeviationDeviation ForecastForecast DeviationDeviation

TonnageTonnage withwith forfor withwith forforQuarterQuarter UnloadedUnloaded = .= .1010 = .= .1010 = .= .5050 = .= .5050

11 180180 175175 55 175175 5511 180180 175175 55 175175 5522 168168 176176 88 178178 101033 159159 175175 1616 173173 141444 175175 173173 22 166166 9944 175175 173173 22 166166 9955 190190 173173 1717 170170 202066 205205 175175 3030 180180 252577 180180 178178 22 193193 131388 182182 178178 44 186186 44

8484 100100

RoundedRounded AbsoluteAbsolute RoundedRounded AbsoluteAbsoluteMAD∑ |deviations|RoundedRounded AbsoluteAbsolute RoundedRounded AbsoluteAbsolute

ActualActual ForecastForecast DeviationDeviation ForecastForecast DeviationDeviationTonageTonage withwith forfor withwith forfor

QuarterQuarter UnloadedUnloaded = .= .1010 = .= .1010 = .= .5050 = .= .5050

11 180180 175175 55 175175 55

MAD = nFor = .1011 180180 175175 55 175175 55

22 168168 176176 88 178178 101033 159159 175175 1616 173173 141444 175175 173173 22 166166 99

= 84/8 = 10.50

For = 5044 175175 173173 22 166166 9955 190190 173173 1717 170170 202066 205205 175175 3030 180180 252577 180180 178178 22 193193 1313

= 100/8 = 12.50For = .50

88 182182 178178 44 186186 448484 100100

RoundedRounded AbsoluteAbsolute RoundedRounded AbsoluteAbsoluteMSE =∑ (forecast errors)2

RoundedRounded AbsoluteAbsolute RoundedRounded AbsoluteAbsoluteActualActual ForecastForecast DeviationDeviation ForecastForecast DeviationDeviationTonageTonage withwith forfor withwith forfor

QuarterQuarter UnloadedUnloaded = .= .1010 = .= .1010 = .= .5050 = .= .5050

11 180180 175175 55 175175 55For = .10

MSE = n

11 180180 175175 55 175175 5522 168168 176176 88 178178 101033 159159 175175 1616 173173 141444 175175 173173 22 166166 99

= 1,558/8 = 194.75

For = 5044 175175 173173 22 166166 9955 190190 173173 1717 170170 202066 205205 175175 3030 180180 252577 180180 178178 22 193193 1313

= 1,612/8 = 201.50For = .50

88 182182 178178 44 186186 448484 100100

MADMAD 1010..5050 1212..5050

RoundedRounded AbsoluteAbsolute RoundedRounded AbsoluteAbsoluteMAPE =100 ∑ |deviationi|/actuali

n

i = 1RoundedRounded AbsoluteAbsolute RoundedRounded AbsoluteAbsoluteActualActual ForecastForecast DeviationDeviation ForecastForecast DeviationDeviationTonageTonage withwith forfor withwith forfor

QuarterQuarter UnloadedUnloaded = .= .1010 = .= .1010 = .= .5050 = .= .5050

11 180180 175175 55 175175 55For = .10

MAPE =n

i 1

11 180180 175175 55 175175 5522 168168 176176 88 178178 101033 159159 175175 1616 173173 141444 175175 173173 22 166166 99

= 45.62/8 = 5.70%

For = 5044 175175 173173 22 166166 9955 190190 173173 1717 170170 202066 205205 175175 3030 180180 252577 180180 178178 22 193193 1313

= 54.8/8 = 6.85%For = .50

88 182182 178178 44 186186 448484 100100

MADMAD 1010..5050 1212..5050MSEMSE 194194..7575 201201..5050

RoundedRounded AbsoluteAbsolute RoundedRounded AbsoluteAbsoluteRoundedRounded AbsoluteAbsolute RoundedRounded AbsoluteAbsoluteActualActual ForecastForecast DeviationDeviation ForecastForecast DeviationDeviation

TonnageTonnage withwith forfor withwith forforQuarterQuarter UnloadedUnloaded = .= .1010 = .= .1010 = .= .5050 = .= .5050

11 180180 175175 55 175175 5511 180180 175175 55 175175 5522 168168 176176 88 178178 101033 159159 175175 1616 173173 141444 175175 173173 22 166166 9944 175175 173173 22 166166 9955 190190 173173 1717 170170 202066 205205 175175 3030 180180 252577 180180 178178 22 193193 131388 182182 178178 44 186186 44

8484 100100MADMAD 1010..5050 1212..5050MSEMSE 194194..7575 201201..5050

MAPEMAPE 55..7070%% 66..8585%%

1) วธสมการเสนตรงสมการทใชในการพยากรณคอ Ft = 421.2 + 33.6tคานวณคา MAD และ MSE ไดขอมลดงตารางคานวณคา MAD และ MSE ไดขอมลดงตาราง

t At Ft et | et | et2

1 450 454 4 4 161 450 454 -4 4 16

2 495 488 7 7 49

3 518 522 -4 4 16

4 563 555 7 7 49

5 584 589 -5 5 25

| | 27 2 155| et |=27 et2=155

4527eMAD

t

4.5

5nMAD

155e 2 315155

neMSE t

2) วธคาเฉลยเคลอนท 2 ชวงเวลา จะไดการคานวณดงตารางดานลาง

t At Ft et | et | et2

1 450 - - - -

2 495 - - - -

3 518 473 45 45 2,025

4 563 507 56 56 3 1364 563 507 56 56 3,136

5 584 541 43 43 1,849

6 - 574 - - -

| et |=144 et2=7,010

48

3144

n

eMAD

t

2 337,23010,7

neMSE2t

ไ สาเหตของการออกนอกการควบคมของการพยากรณมไดหลายกรณ

◦ วธการพยากรณทใชไมเหมาะสมกบลกษณะขอมล เนองจาก◦ วธการพยากรณทใชไมเหมาะสมกบลกษณะขอมล เนองจาก ไมไดรวมตวแปรสาคญทมผลกบความตองการสนคาเขาไปในวธการพยากรณ

การเปลยนแปลงของตวแปรทวธการพยากรณนนไมครอบคลม เกดตวแปรใหมขน เชน คแขงใหม

◦ เกดการเปลยนแปลงอยางกะทนหน เชน การเกดภยธรรมชาต ◦ เกดการเปลยนแปลงอยางกะทนหน เชน การเกดภยธรรมชาต สงคราม โรคระบาด จะมผลกระทบกบลกษณะความตองการสนคา

◦ วธการพยากรณถกนาไปใชอยางไมถกตอง หรอแปลความหมายจากผลการพยากรณไมถกตอง

สญญาณตดตาม (Tracking Signal)แผนภมควบคม

+ แผนภมควบคม

(Control Chart)

พลาด

UCLออกนอกการควบคม

0

-

ความผดพ

LCL

Measures how well Measures how well forecast is predicting actual valuesR ti f i f

t MADRSFETS

Ratio of running sum of forecast errors (RSFE) to mean absolute deviation (MAD) te(MAD)◦ Good tracking signal

has low valuesSho ld be ithin pper

tMAD

Should be within upper and lower control limits

t A Fเดอนท, t At Ft et

1 47 43 4

2 51 44 7

3 54 50 4

4 55 51 4

5 49 54 55 49 54 -5

6 46 48 -2

7 38 46 -8

8 32 44 -12

9 25 35 -10

10 24 26 210 24 26 -2

t (1)

n At Ft et

1 47 43 41 47 43 4

2 51 44 7

3 54 50 4

4 55 51 4

5 49 54 -5

6 46 48 -2

7 38 46 -8

8 32 44 -12

9 25 35 -109 25 35 10

10 24 26 -2

t (1) (2)

n At Ft et | et |

1 47 43 4 41 47 43 4 4

2 51 44 7 7

3 54 50 4 4

4 55 51 4 4

5 49 54 -5 5

6 46 48 -2 2

7 38 46 -8 8

8 32 44 -12 12

9 25 35 -10 109 25 35 10 10

10 24 26 -2 2

t (1) (2) (3)

n At Ft et | et | et

1 47 43 4 4 41 47 43 4 4 4

2 51 44 7 7 11

3 54 50 4 4 15

4 55 51 4 4 19

5 49 54 -5 5 14

6 46 48 -2 2 12

7 38 46 -8 8 4

8 32 44 -12 12 -8

9 25 35 -10 10 -189 25 35 10 10 18

10 24 26 -2 2 -20

t (1) (2) (3) (4)

n At Ft et | et | et | et |

1 47 43 4 4 4 41 47 43 4 4 4 4

2 51 44 7 7 11 11

3 54 50 4 4 15 15

4 55 51 4 4 19 19

5 49 54 -5 5 14 24

6 46 48 -2 2 12 26

7 38 46 -8 8 4 34

8 32 44 -12 12 -8 46

9 25 35 -10 10 -18 569 25 35 10 10 18 56

10 24 26 -2 2 -20 58

t (1) (2) (3) (4) (5)

n At Ft et | et | et | et |MADt=| et |/n

1 47 43 4 4 4 4 4.00

2 51 44 7 7 11 11 5.50

3 54 50 4 4 15 15 5.00

4 55 51 4 4 19 19 4.75

5 49 54 -5 5 14 24 4.80

6 46 48 -2 2 12 26 4.336 46 48 2 2 12 26 4.33

7 38 46 -8 8 4 34 4.86

8 32 44 -12 12 -8 46 5.75

9 25 35 10 10 18 56 6 229 25 35 -10 10 -18 56 6.22

10 24 26 -2 2 -20 58 5.80

t (1) (2) (3) (4) (5) (6)

n At Ft et | et | et | et |MADt=| et |/n

TSt= et/MADt

1 47 43 4 4 4 4 4.00 1.00

2 51 44 7 7 11 11 5.50 2.00

3 54 50 4 4 15 15 5.00 3.00

4 55 51 4 4 19 19 4.75 4.00

5 49 54 -5 5 14 24 4.80 2.92

6 46 48 -2 2 12 26 4.33 2.776 46 48 2 2 12 26 4.33 2.77

7 38 46 -8 8 4 34 4.86 0.82

8 32 44 -12 12 -8 46 5.75 -1.39

9 25 35 10 10 18 56 6 22 2 899 25 35 -10 10 -18 56 6.22 -2.89

10 24 26 -2 2 -20 58 5.80 -3.45

4.00

5.00

2.003.00

1 000.001.00

-3.00-2.00

-1.00 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

-5.00-4.00

ปจจยทสาคญม 2 ปจจย◦ ความแมนยาของการพยากรณ ◦ คาใชจายในการพยากรณ ◦ คาใชจายในการพยากรณ

ปจจยยอยอนๆ◦ จานวนขอมลในอดตมากนอยเพยงใด

ไ◦ มคอมพวเตอรชวยประมวลผลหรอไม◦ ผพยากรณมความสามารถในการใชเทคนคการพยากรณมากนอยเพยงใด

◦ ระยะเวลาในการเกบขอมล วธคาเฉลยเคลอนท และ วธเอกซโปเนนเชยลปรบเรยบ เหมาะกบการพยากรณระยะส น เพราะผลลพธจากการคานวณจะพยากรณเพยง ชวงเวลาถดไปเทานน

สมการเสนตรงสามารถใชพยากรณในระยะยาวได การพยากรณเชงคณภาพหลายวธสามารถใชพยากรณระยะยาวได เพราะไมตองใชขอมลในอดต เชน ความเหนของผบรหาร และ วธเดลฟาย

ปจจยระยะเวลาในการพยากรณ

ระยะส น ระยะปานกลาง ระยะยาว

ความถในการพยากรณ

บอย บางคร ง ไมบอย

ประเภทสนคาทพยากรณ

1 ประเภท กลมสนคาประเภทเดยวกน

ทกประเภทรวมกน

ตวแบบการพยากรณ การปรบเรยบ สมการเสนตรง แลวแตผบรหารสมการเสนตรงการวเคราะหเชงถดถอย

ฤดกาลการวเคราะหเชงถดถอย

บทบาทผบรหาร นอย ปานกลาง มาก

คาใชจายในการพยากรณ

ตา ปานกลาง สง

ระยะเวลา

ความแมนยา

ความเชอถอได ความเชอถอได

หนวยวดเหมาะสม

เกบเปนเอกสาร

งาย งาย